Air Quality Reports in South Portland

What does it smell like?

When users report a smell, they are given the option of providing a description for what the smell they are reporting. In order to get a better idea of what kinds of stinks are plaguing the residents of Portland, we counted the number of times people used specific words in their descriptions, and then categorized them into six groups.

The cart above shows a clear trend regarding smell reports: the vast majority of smells that motivate people to file a report originate from the petrochemical industry.

How were smell types calculated?

To categorize what types of smells people reported, we counted the words used in each smell report description. First, each description was split into the descriptions component words. Then, words unrelated to smell descriptions, such as "always", "very", and "ruining", were removed. Next, a list of synonyms were used to remove spelling errors, differences in tense and plurality, and regional and cultural differences. For example: "sewer" was replaced with "sewage", "gassy" was replaced with "gas", "pavement" with "asphalt", and "eggs" with "sulfur".

At this point, we have the un-grouped terms used in each description. When we count these terms, we can create the chart below, which shows terms used in descriptions which were counted at least fifty times. Several descriptions of the local smells are repeated often: "petroleum", "tar", "oil", "asphalt", "sulfur", "gasoline". Complaints also mention feeling nausea, throat irritation, and headaches.

These terms were categorized by where they most likely originated, except for terms that only related to safety ("carcinogen", "nausea", etc.). For example: "oil", "gas", and "propage" were all categorized as "petrochemical"; "rubber", "asphalt", and "tires" were mapped to "transportation"; and "carcinogen", "electrical", and "bitter" were mapped to "unknown, dangerous".

To see exactly how these charts were created see this observable notebook.

Methodology

The data used in this visualization is obtained through an API call. Since this is a public API, SmellMyCity has taken certain precautions to protect personal information which could impact the reliability of the data. In particular:

• Each submission is treated as a unique data point; we cannot identify multiple submissions from the same user.
• Location data are skewed within a short radius to protect actual locations.
• The data are contributed by users and must be treated similarly to public-facing surveys, as opposed to purely objective measures.

Prevailing wind direction is obtained separately from a NOAA database. The direction associated with each data point is the prevailing direction on the day of the report, but may not represent wind conditions at the particular moment the report was submitted.

Finally, the text analysis presented above required many design choices from the researchers, including determining the categories in which to group comments and developing a list of synonyms.